devwithjevJev-Powered Browser Agent Benchmark
The benchmark uses Jev to compare two browser agents with different models on task completion efficiency, scoring correctness, time, actions, and cost.
Also filed under Agents & browsers
- Jev-kit for Claude Code
Cadaverr’s Jev-kit integrates TypeSafe’s Jev model into Claude Code for tasks including tool-call guarding, sub-agent sizing, file search, and browser automation. Jev answers pick-one and yes/no questions in about 0.3s, according to the author.
- Jev Browser Automation
jev-browser takes a task and URL, then uses Jev to choose one action per step from the page’s clickable, typeable, and selectable elements. It scores whether the goal is met or the run is stuck.
- Jev Gates Agent Memory Saves
The author used Jev to check whether proposed agent memories were supported by the original conversation and guide whether to save, skip, or defer them. On 100 synthetic cases at a .40 cutoff, Jev kept 39 of 50 memories labeled worth keeping and saved no candidates labeled skip or defer; its median gate latency was 250 ms.
- Jev in the Knocklabs Agent Harness
Chris Bell says he implemented this in the Knocklabs agent harness using Jev, based on common types of requests seen in live traces.